letters

The 2.0 Å crystal structureof catalase-peroxidasefrom HaloarculamarismortuiYusuke Yamada1, Taketomo Fujiwara2, Takao Sato1,Noriyuki Igarashi3 and Nobuo Tanaka1

1Department of Life Science, Graduate School of Bioscience andBiotechnology, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku,Yokohama 226-8501, Japan. 2Department of Biology and Geosciences,Faculty of Science, Shizuoka University, 836 Ohya, Shizuoka 422-8529,Japan. 3Institute of Materials Structure Science, High Energy AcceleratorResearch Organization, 1-1 Oho, Tsukuba 305-0801, Japan.

Published online: 12 August 2002, doi:10.1038/nsb834

Catalase-peroxidase is a member of the class I peroxidasesuperfamily. The enzyme exhibits both catalase and peroxi-dase activities to remove the harmful peroxide molecule fromthe living cell. The 2.0 Å crystal structure of the catalase-peroxidase from Haloarcula marismortui (HmCP) reveals thatthe enzyme is a dimer of two identical subunits. Each subunitis composed of two structurally homologous domains with atopology similar to that of class I peroxidase. The active site ofHmCP is in the N-terminal domain. Although the arrange-ment of the catalytic residues and the cofactor heme b in theactive site is virtually identical to that of class I peroxidases, theheme moiety is buried inside the domain, similar to that in atypical catalase. In the vicinity of the active site, novel covalentbonds are formed among the side chains of three residues,including that of a tryptophan on the distal side of the heme.Together with the C-terminal domain, these covalent bonds fixtwo long loops on the surface of the enzyme that cover the sub-strate access channel to the active site. These features providean explanation for the dual activities of this enzyme.

Catalase-peroxidases are present in many microorganisms.These enzymes exhibit both catalase and peroxidase activities,

which defend against oxygen toxicity by removing H2O2 fromthe cell. Catalase converts two equivalents of hydrogen peroxideinto water and oxygen via a two-step reaction cycle in whichH2O2 alternately oxidizes and reduces the heme iron at the activesite. For peroxidase, oxidation of the heme iron involves a H2O2

molecule, similar to that in the catalase-catalyzed reaction.Reduction of the heme iron, however, involves hydrogen donorssuch as NADH, not a second H2O2 molecule.

The catalase-peroxidases that have been characterized bio-chemically are either homodimers or homotetramers1. A typicalsubunit is ∼ 80 kDa in molecular mass, with a single heme bcofactor at its active site. The primary structure of the subunitcan be divided into two halves that have significant sequencesimilarity, presumably arising from a gene duplication event2.Both halves also share sequence similarity with yeastcytochrome c peroxidase (CCP) and eukaryotic ascorbate per-oxidase (APX), which are members of the class I peroxidasesuperfamily3. In contrast, they share no sequence similaritywith typical monofunctional catalases, such as bovine liver cata-lase4 and Escherichia coli hydroperoxidase II5. A heme-bindingconsensus sequence is present only in the N-terminal half ofeach subunit and thus likely contains the active site. The C-ter-minal half does not contain a heme-binding motif and its func-tion is unclear.

A catalase-peroxidase from Haloarcula marismortui (HmCP)was reported to be a bifunctional enzyme6 and has a highsequence similarity with other catalase-peroxidases (40–90%)1.Here we report the crystal structure of the HmCP at 2.0 Å; this isthe first report for the three-dimensional structure of any cata-lase-peroxidase.

Overall structureThe crystal structure of HmCP was determined using the MADmethod with a K2PtCl4 derivative. An asymmetric unit of theHmCP crystal contains two identical subunits related by non-crystallographic two-fold symmetry (Fig. 1a,b). Each subunitbinds one heme b and ∼ 700 solvent molecules. Residues 1–17 ofboth subunits and residues 295–301 of subunit 1 are not included in the final model because they are invisible in the elec-tron density map; these segments are presumably very flexible inthe molecule. The sequence corresponding to the N-terminal

nature structural biology • volume 9 number 9 • september 2002 691

Fig. 1 Overall structure of HmCP. a, Two views of the dimer related by a 90° rotation. The N- and C-terminal domains of each subunit 1 are purpleand blue, respectively, whereas those in subunit 2 are yellow and gold, respectively. The noncrystallographic two-fold symmetry in the dimer is rep-resented by a black ellipse. Heme b is shown as a red stick model. b, The N- and C-terminal domains of HmCP have homologous structures. The 10 helices in each domain are labeled according to the naming scheme of CCP and APX. The black ellipse represents the pseudo two-fold rotationsymmetry axis relating the two domains.

ba

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

disordered region is not conserved in catalase-peroxidases7;thus, this segment is unlikely to be important for the function ofthe enzyme. Few significant differences exist between the struc-tures of the two subunits; the r.m.s. deviations between the sub-units are 0.25 Å for Cα and 0.63 Å for all atoms.

The structure of HmCP contains mostly α-helices with few β-sheets. Consistent with predictions from sequence analysis2,each subunit is organized into two structurally similar domains(Fig. 2b): the N-terminal domain (HmCP-N, residues 18–430)and the C-terminal domain (HmCP-C, residues 431–731). Thetopological arrangements of secondary structural elements in

the two domains are identical to thoseof CCP8 and APX9, which belong to theclass I peroxidase superfamily3. Usingstructure-based sequence alignment ofHmCP-N, HmCP-C, CCP and APX(Fig. 2a), 10 helices from each domainare designated A–J, according to thenaming scheme of CCP and APX.HmCP-N and HmCP-C share 30.2%and 42.7% sequence similarity, respec-tively, with CCP.

HmCP-N contains the heme b cofac-tor at the active site. The overall struc-ture of HmCP-N superimposes wellwith that of CCP after fitting theresidues within the secondary struc-tures (r.m.s. deviation of 1.07 Å for Cαatoms; Fig. 2b), except for several inser-tions, including the long extended N-terminal end (N-ter, residues 18–48),a short stretch (SS, residues 74–83) andthree large loops (LL1, residues182–232; LL2, residues 274–302; andLL3, residues 347–364). All of theseinsertions are on the surface of thedomain. Notably, LL1 is at one edge ofthe heme, whereas LL2 is opposite LL1at the other edge of the heme and inter-acts with HmCP-C of the other subunitin the dimer. LL3 is far from the activesite; therefore, its role in the enzyme isnot well defined.

The heme is buried inside HmCP-N,and substrate access to the active site isthrough a narrow channel that preventsaccess of a large substrate (Fig. 3a). The

side chains of Glu 194 of LL1and Glu 271 of LL2 are at the sur-face of the molecule and mark the acidic entrance of the channel(Fig. 3a). The heme moiety is ∼ 20 Å from Glu 271; the arrange-ment of a buried heme active site is similar to that in mono-functional catalases3 but is different from that in peroxidases.The cylindrical channel around the heme iron is ∼ 3.5 Å indiameter at Ser 305 (calculated using VOIDOO10), and thewidth of the channel increases from Ser 305 to the molecularsurface. Mutation of the corresponding Ser (Ser 315) to a Thr in the catalase-peroxidase from Mycobacterium tuberculosis(KatG) results in a loss of the activation to the anti-tuberculosis

692 nature structural biology • volume 9 number 9 • september 2002

a

b

Fig. 2 Structural comparison among class Iperoxidase superfamily. a, Structural-basedamino acid sequence alignment. Secondarystructure of HmCP-N and HmCP-C is shownabove the sequence alignment. Dotted linesrepresent insertions in HmCP-N. Conservedresidues are in the small blue boxes; identicalresidues among all four sequences are indi-cated by white letters on red background;and similar residues are indicated by red let-ters. The regions in the large black boxesexhibit <2.5 Å r.m.s. deviations for Cα in allstructures, which were calculated whenstructures were superimposed with eachother after fitting the residues within thesecondary structures. b, The Cα trace repre-sentations of HmCP-N (green) and CCP (blue)are superimposed. The long insertions inHmCP-N (red) are labeled.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

drug (INH, an aniline derivative) with no loss of either peroxi-dase or catalase activity11. Based on the structure of HmCP,Ser 315 in KatG is likely situated along the access channel.Replacing the side chain of Ser with a bulkier Thr might intro-duce steric hindrance and therefore prevent INH from accessingthe distal heme pockets.

The overall structure of HmCP-C is also similar to CCP (r.m.s.deviation of 1.35 Å for Cα atoms in helices and strands). Becauseit does not contain a heme moiety, the space corresponding tothe heme pocket is filled with two short β-strands (betweenhelices Dc′ and E) and a loop (between helices F and G) (Fig. 1b).HmCP-C has no large insertions to cover the molecular surfaceas found in HmCP-N. HmCP-N and HmCP-C are related by apseudo two-fold symmetry in the molecule (Fig. 1b) and areconnected by a long loop that does not have any defined sec-ondary structural element.

The active siteThe organization of the active site of HmCP-N is similar tothose of CCP8 and APX9. The r.m.s. differences for atoms in theheme b moiety and eight conserved residues in the vicinity are0.27 Å between CCP and HmCP-N, and 0.29 Å between APXand HmCP-N. Some differences do exist. Notably, a potassiumion near the Trp at the proximal side of heme b in APX isreplaced by a water molecule in HmCP and CCP. This potassi-um ion was suggested to be essential for the formation of areaction intermediate called compound I, a porphyrin π-cationradical Fe(IV)=O species, in APX9. A water molecule near theproximal Asp residue that is conserved in APX is also absent inHmCP and CCP. The electrostatic environment of HmCP inthe vicinity of proximal His 259, Asp 372 and Trp 311 does notresemble that of APX but is similar to that of CCP. In CCP, thesecond oxidizing equivalent of compound I is known to be atryptophan radical12. These observations indicate that the com-pound I intermediate of HmCP and possibly other catalase-

peroxidases13 would be formed via a different mechanism fromAPX.

Novel covalent modificationIn the vicinity of the active site of HmCP, continuous electrondensity is observed between Cε1 of Tyr 218 and Cη2 of Trp 95,and between Cε2 of Tyr 218 and Sδ of Met 244 (Fig. 3b) in bothsubunits of the dimer. Regardless of whether the model wasrefined with or without steric restraint among the three sidechains, the electron density map indicates the existence of cova-lent bonds between these side chains. The distances between thecovalently bonded atoms in the two subunits are Tyr 218Cε1–Trp 95 Cη2, 1.68 (±0.02) Å; and Tyr 218 Cε2–Met 244 Sδ,1.72 (±0.03) Å. These novel covalent bonds have not beenreported previously.

To investigate whether these covalent bonds are also present inthe enzyme in solution, we performed tryptic digestion ofHmCP followed by mass spectrometry analysis of the reactionproducts. The presence of a species that corresponds to thecrosslinked product of two peptide fragments (residues 93–102and 198–235) would confirm that the covalent bonds are indeedformed. We were unable to definitively prove the existence of thecovalent bonds in the enzyme in solution because we did notidentify any fragments corresponding to residues 93–102,198–235 or the crosslinked species between these peptides.However, this result does not rule out the possibility that themodified fragment may have anomalous chromatographicproperties and may be trapped on the gel filtration columnbefore mass spectrometry. Consistent with this hypothesis, massspectrometry analysis of cyanobromide-cleaved HmCP also didnot detect any peptide fragments containing Trp 95, Tyr 218 andMet 244 (ref. 1).

The side chain of Met 244 exists as a sulfonium ion with a pos-itive charge because Sδ of Met 244 donates a pair of electrons toform the extra covalent bond. The guanidino group of Arg 409

nature structural biology • volume 9 number 9 • september 2002 693

a b

Fig. 3 Active site structure. a, The access channel to the active site. Theside of the surface toward the solvent is red, and that facing the interiorof the protein is blue. b, Stereo view of the 2Fo – Fc omit maps for theregion of Trp 95, Tyr 218 and Met 244 contoured at the 2.0 σ level. Thecolor coding in the stick model is red, oxygen; blue, nitrogen; yellow, car-bon; and green, sulfur. c, The environment of Arg 409. Red and orangespheres represent water molecules and iron atom of the heme, respec-tively. Hydrogen bonds are represented as green dotted lines. Distances(Å) between the hydrogen bonding atoms are labeled.

c

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

forms hydrogen bonds with Oη of Tyr 218 and the amide nitro-gen atom of Met 244; these interactions stabilize the covalentbond between Tyr 218 and Met 244 by fixing the relative posi-tions of these residues (Fig. 3d). Tyr 218 anchors LL1 to the mol-ecular surface by forming covalent bonds with Trp 95 andMet 244 at the distal side of heme b (Fig. 3b). The covalent bondbetween Trp 95 and Tyr 218 does not perturb the relative orien-tation of Trp 95 and the heme, which is important in reaction ofthe second H2O2 with compound I14. The conservation of Trp 95,Tyr 218, Met 244 and Arg 409 suggests that the novel covalentbonds observed in HmCP may be common in other catalase-peroxidases. Furthermore, the covalent bonds may be importantfor the catalase activity of catalase-peroxidase because only thecatalase activity is lost upon mutation of Trp 95 (refs 7,14). Toinvestigate the functional role of the covalent bonds, we are cur-rently studying the effects of mutating Tyr 218 and Met 244 onthe activities of HmCP.

HmCP-C interacts with both LL1 and LL2 of HmCP-N thatpartially cover the surface entry to the access channel at the distalside of the heme. The aliphatic portion of Gln 286 in LL2 formstight interactions with hydrophobic residues in HmCP-C.Although H2O2 could potentially access the buried active site ofHmCP without conformational changes in both LL1 and LL2,access of large substrates such as NADH to the active site mustinvolve movement of these two loops. Movement of the C-termi-nal domain coupled to the two long loops represents an attrac-tive explanation for the dual activities of catalase-peroxidases.The flexible linker between HmCP-N and HmCP-C and theobservation that several mutations in the C-terminal domain of

KatG (the catalase-peroxidase from M. tuberculosis) abolishenzymatic activity15 are consistent with this hypothesis.

Solvent ionsThe halophilic HmCP requires a high ionic strength for its stabil-ity. The crystal structure of HmCP reveals that its molecular sur-face contains more acidic (Glu and Asp) (54%) and less basic(Arg, Lys and His) (8%) residues than non-halophilic catalase-peroxidases. Of the ions observed in the crystal structure, six sul-fate ions bind basic side chains and form hydrogen bonds witheither main chain amides or water molecules. Only 2 of the 16chloride ions are at the interface between the adjacent moleculesrelated by the crystallographic symmetry; the remaining 14chloride ions and 6 chloride or potassium ions are either insidethe molecule or at the dimer interface. These ions form saltbridges with basic side chains, or hydrogen bonds to amidegroups or water molecules. Therefore, these interactions may beessential to maintain the active form of the enzyme. Consistentwith this, the purified enzyme is degraded without ions in solu-tion (data not shown).

MethodsCrystallization and data collection. HmCP was purified from H. marismortui and crystallized as described16. HmCP was crystal-lized with the hanging-drop vapor diffusion method at 20 °C for1–2 weeks. Protein solution containing 40–50 mg ml–1 HmCP in10 mM Tris-HCl, pH 8.0, 2.0 M (NH4)2SO4 and 0.5 M KCl was equili-brated against 1.0 ml of 3.0–3.2 M (NH4)2SO4. A heavy atom deriva-tive was prepared by soaking crystals in the mother liquorcontaining 2 mM K2PtCl4 for 1 d. Crystals were flash frozen at 100 K

694 nature structural biology • volume 9 number 9 • september 2002

Table 1 Data collection, phasing and refinement statistics

Data collectionNative K2PtCl4

peak edge remoteWavelength (Å) 1.0000 1.0718 1.0722 0.9800Resolution (Å)1 34.1–2.0 (2.11–2.00) 36.5–2.2 (2.32–2.20) 36.5–2.2 (2.32–2.20) 36.5–2.2 (2.32–2.20)Reflections

Observed1 479,060 (67,473) 369,271 (53,498) 369,776 (53,577) 370,509 (53,770)Unique1 128,391 (18,662) 97,380 (14,029) 97,496 (14,049) 97,585 (14,081)

Completeness (%)1 99.9 (99.7) 99.0 (98.4) 99.0 (98.3) 99.2 (98.7)I / σ(I)1 9.3 (5.0) 13.4 (2.6) 14.0 (3.1) 11.9 (2.4)Rmerge (%)1 5.2 (14.0) 4.6 (24.9) 4.3 (21.3) 5.3 (29.1)

Phasing2

Phasing power 0.36 / 0.46 – / – 0.62 / 0.74RCullis

Isomorphous 0.92 / 0.94 – / – 0.83 / 0.88Anomalous – / 0.79 – / – – / 0.88

Refinement statisticsResolution (Å) 34.1–2.0 Average B-factor (Å2)Total atoms 12,606 ProteinRwork / Rfree (%) 18.5 / 20.5 Main chain 23.4R.m.s. deviations Side chain 25.6

Lengths (Å) 0.0073 Heme bAngles (°) 1.62 Iron 19,1

Porphyrin 18.0Sulfate ions 50.6Chloride ions 25.4Water 29.7

1Values in parentheses are for the highest resolution shell.2Values are listed as centric / acentric.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

after soaking for 30 min in 100 mM Tris-HCl, pH 8.0, 2.5 M (NH4)2SO4,0.5 M KCl with 1.0 M Li2SO4 as cryoprotectants. MAD data from thederivatized crystals were collected at BL18B (Photon Factory) atthree wavelengths: one near the LIII absorption edge (λ2 = 1.0722 Å,maximum of |f′|) and one near the peak (λ3 = 1.0718 Å, maximum off′′ ) of platinum, and one at a high-energy remote wavelength (λ1 =0.9800 Å). The native data set was collected at BL6A (PhotonFactory) and BL40B2 (Spring-8). The platinum derivative and nativedata sets were processed with MOSFLM/SCALA17,18.

Structure determination and refinement. Three platinumatoms were found from the anomalous difference Patterson map.The refinement of the heavy atom parameters and phasing wereperformed at 2.5 Å resolution with MLPHARE19. Self-rotation func-tion20 and these heavy atom positions were used to identify theorientation and position of the noncrystallographic two-fold axis.After density modification and phase extension to 2.2 Å includingsolvent flattening, histogram matching and two-fold averagingusing DM21, the electron density map became clear enough to beinterpreted. O22 was used to examine the maps and to build molec-ular models. The refinement was carried out with CNS.

After the R-factor of the MAD structure dropped to 24%, thenative structure was refined using the MAD structure as a startingmodel. A rigid body refinement and iterative refinements with O22

and CNS23 were followed. Noncrystallographic symmetry restraintswere not employed. The final model consists of amino acids 18–294and 302–731 of subunit 1 and 118–731 of subunit 2, 6 sulfate ions,16 chloride ions, 6 chloride or potassium ions and 1,390 water mole-cules. The types of solvent molecules were determined from elec-tron densities and their environments. Crystallographic statistics aresummarized in Table 1.

Structure Analysis. Analysis of the stereochemical quality of theprotein model and assignment of secondary structure were con-ducted with PROCHECK24. Superpositions of class I peroxidase fami-ly members were done with the Magic Fit command of Swiss-PdbViewer25. Figures were prepared with ESPript26, MolScript27,BobScript28, Raster3D29 and GRASP30.

Coordinates. Coordinates have been deposited in the Protein DataBank (accession code 1ITK).

AcknowledgmentsThe present research was undertaken with the approval of the Photon FactoryAdvisory Committee, Japan, and the Japan Synchrotron Radiation Research

Institute (JASRI). The authors wish to express their thanks to the staff at thePhoton factory and SPring-8 for their help and the use of the diffractometer. Theproject was partly supported by Grants-in-Aid for Scientific Research from theMinistry of Education, Culture, Sports, Science and Technology of Japan; theACT-JST Program, Japan Science and Technology Corporation; and research grantfrom Salt-Science Foundation.

Competing interests statementThe authors declare that they have no competing financial interests.

Correspondence should be addressed to N.T. email: ntanaka@bio.titech.ac.jp

Received 14 February, 2002; accepted 23 July, 2002.

1. Cannac-Caffrey,V. et al. Biochimie 80, 1003–1011 (1998).2. Zamocky, M., Janecek, S. & Koller, F. Gene 256, 169–182 (2000).3. Welinder, K.G. Biochim. Biophys. Acta 1080, 215–220 (1992).4. Murthy, M.R.N., Reid, T.J. III, Sicignano, A., Tanaka, N. & Rossmann, M.G. J. Mol.

Biol. 152, 465–499 (1981).5. Bravo, J. et al. Structure 3, 491–502 (1995).6. Cendrin, F., Jouve, H. M., Gaillard, J., Thibault, P. & Zaccai, G. Biochim. Biophys.

Acta 1209,1–9 (1994)7. Zamocky, M., Regelsberer, G., Jakopitsch, C. & Obinger C. FEBS Lett. 492,

177–182 (2001).8. Finzel, B.C., Poulos, T.L. & Kraut, J. J. Biol. Chem. 259, 13027–13036 (1984).9. Patterson, W.R. & Poulos, T.L. Biochemistry 34, 4331–4341 (1995).

10. Kleywegt, G.J. & Jones, T.A. Acta Crystallogr. D 50, 178–185 (1994).11. Wengenack, N.L. et al. J. Infect. Dis. 176, 722–727 (1997).12. Sivaraja, M., Goodin, D.B., Smith, M. & Hoffman, B.M. Science. 245,738–740

(1989).13. Chouchane, S., Lippai, I. & Magliozzo R.S. Biochemistry 39, 9975–9983 (2000)14. Hillar, A. et al. Biochemistry 39, 5868–5875 (2000).15. Ramaswamy, S. & Musser, J.M. Tuber. Lung Dis. 79, 3–29 (1998).16. Yamada, Y. et al. Acta Crystallogr. D 57, 1157–1158 (2001).17. Leslie, A.G.W. Proceedings of the CCP4 study weekend (eds Sawyer, L., Isaacs,

N. & Bailey, S.) 44–51 (SERC Daresbury Laboratory, Warrington; 1993).18. Evans, P.R. Proceedings of the CCP4 study weekend (eds Wilson, K.S., Davies, G.,

Ashton, A.W. & Bailey, S.) 97–102 (SERC Daresbury Laboratory, Warrington;1997).

19. Otwinowski, Z. Proceedings of the CCP4 study weekend (eds Wolf, W., Evans,P.R. & Leslie, A.G.W.) 80–86 (SERC Daresbury Laboratory, Warrington; 1991).

20. Tanaka, N. Acta Crystallogr. A 33, 191–193 (1977).21. Cowtan, K. Joint CCP4 and ESF-EACBM Newsletter on Protein Crystallography

31, 34–38 (1994).22. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. Acta Crystallogr. A 47,

110–119 (1991).23. Brünger, A.T. et al. Acta Crystallogr. D 54, 905–921 (1998).24. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. J. Appl.

Crystallogr. 26, 283–291 (1993).25. Guex, N. & Peitsch, M.C. Electrophoresis 18, 2714–2723 (1997).26. Gouet, P., Courcelle, E., Stuart, D.I. & Metoz, F. Bioinformatics 15, 305–308

(1999).27. Kraulis, P.J. J. Appl. Crystallogr. 24, 946–950 (1991).28. Esnouf, R.M. J. Mol. Graph. Model. 15, 132–134 (1997).29. Merritt, E.A. & Bacon, D.J. Methods Enzymol. 277, 505–524 (1997).30. Nicholls, A., Sharp, K. A. & Honig, B. Proteins 11, 281–282 (1991).

nature structural biology • volume 9 number 9 • september 2002 695

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y